►
From YouTube: DASH Behavioral Model WG Dec 15 2022
Description
We have tests marked 'don't run' in BMV2
Big challenge - not all of us are P4 savvy, if we had chosen C there may have been a bigger universe to work on it
Simulator, Compiler, etc...
Open Source DPDK backend for the compiler (Andy)
Here we have not done scripts that take SAI APIs down to 'something' in the P4 lang
P4 DPDK support is happening, in a few months we might be able to use it better from a DASH standpoint w/features in P4 DPDK
Need BMV2 owner; knowledge and expertise is not widespread.
Gerald to check w/Engineering (via Deepak and MSR) for resources.
A
Thank
you
for
coming
to
dash
behavioral
model
meeting
for
today,
the
last
one
before
the
December
holidays,
December
15th
I'll
go
ahead
and
turn
off
my
camera.
So
what
I'm
going
to
project
right
now
is
just
my
screen
to
show
the
GitHub
behavioral
model
project.
This
link
I,
usually
included
in
the
meeting
notes
every
week
somewhere
near
the
bottom
of
the
meeting,
notes
if
it's
something
you'd
like
to
go
look
at
on
your
own.
But
what
we
have
here
is
the
all
up
behavioral
model
here
and
the
backlog
here.
A
The
items
that
we've
done
you
know
here
are
under
the
done
status,
honey,
there's
quite
a
few
things
that
we've
gotten
done
in
the
last
nine
months
or
so,
and
then
we
also
had
a
few
bugs
that
we
figured
out,
but
I
think
you
know
I
think
the
frustrating
part
of
the
mv2
is
that
we
have
a
lot
of
tests
that
we're
writing
that
just
don't
not
seem
to
run
in
BMV
too.
We
don't
have
the
the
support
that
we
need.
Would
that
be
a
true
statement,
Chris
or
rest
of
the
team.
B
Yeah,
that's
true
we're
starting
to
write
more
and
more
tests
that
have
to
be
marked,
don't
run
bmv2
and
it's
almost
the
converse
of
let's
say
the
vision
that
we
set
out
to
have
a
model
that
says
this
is
how
it
shall
work
now.
Go
and
Implement
we're
working
a
little
bit
upside
down
from
that
where
the
vendor
implementations
are
actually
starting
to
work
even
with
libsi.
B
A
C
A
C
D
A
Right,
I
I
think
there
are
a
few
people
picking
up
I
mean
James
is
pretty
interested
in
picking
up
items
and
I've
tried
I've
been
trying
to
find
people
internally
at
Microsoft
like
over
in
MSR
Microsoft
research
group,
or
you
know,
PhD
candidates
or
you
know,
I've
talked
to
someone
I
know
over
at
Oracle
to
try
and
find
you
know
different
resources,
so
I've
been
looking
at
different
options.
I
haven't
been
super
successful,
but
you
know
that's
those
are
the
things
I've
been
kind
of
looking
at
doing.
C
So
I
mean
I
have
a
little
bit
of
history
with
attempting
to
get
like
grad
students,
and
things
like
that
interested
in.
Like
finishing
the
implementation
of
the
portable
switch
architecture
on
bnb2
and
I,
was
like
you
had
one
or
two
grad
students
interested
in
doing
stuff
like
that
and
for
a
year
or
so,
and
it
only
got
partially
done
so
I
mean
it
most
most
popular
I,
I,
don't
know
what
I
I
think
this
is
true
of
Linux
and
I.
Think
that's
true
and
I
did
a
survey
of
p4.org
projects.
E
A
C
B
B
A
D
I'm
not
saying
we
should
switch
at
this
point,
but
I
think
that
that
that's
like
a
one
of
the
challenges
and
I
think
like
some
of
these
things
like
take
the
ACL,
for
example
like
in
C,
it
could
have
just
been
modeled
as
just
a
sequential
or
walk
through
a
list
of
rules
and
from
a
behavioral
standpoint.
That's
perfectly
valid
right.
I.
C
D
Yeah,
no
so
yeah
I
think
it's
it's
I
think
this
is
a
a
challenge
with
with
like
p4s
meaning
the
behavioral
model.
Is
that,
like
there,
you
know,
I,
don't
know
in
different
organizations
but
like
here
we
we,
we
just
don't
have
like
a
lot
of
expertise
in
P4,
the
simulator
or
you
know
where.
Obviously
we
look
at
the
P4
to
understand
the
behavioral
model,
but
it's
like
it.
That's
a
challenge.
C
They,
what
they
have
not
done,
which
Mari
and
I
think
and
team
have
done,
is
creating
these
kind
of
scripts
that
can
take
SCI
apis
and
convert
them
into
some
kind
of
BMV.
Twos
I,
don't
know,
actually
know
that
you
guys
know
better
than
I.
Do
probably
they've
looked
at
it.
What
that
code
does
Texas
IPA
is
down
to
Something
in
the
people,
land
and
I.
Don't
know
if
that's
easily
portable
to
DBK
or
not.
F
Right
I
didn't
purposely
mention
it
earlier,
but
you
know
definitely
the
dpdk
support
the
people.
Dpdk
support
for
some
of
this
feature
is
happening,
and
now
we
have
more
traction
there
than
before.
So
in
a
few
months,
I
think
we'll
be
able
to
use
it
better
in
from
a
dash
standpoint,
with
the
features
that
are
available
in
the
p4dpdk
and.
C
C
Talk
offline
but
there's
definitely
things
there's
gaps
for
dpdk
as
well,
that
are
specific
to
that.
That
are
functional
reference
model
only
and
probably
would
not
be.
Nobody
would
want
to
use
a
sequential
going
through
a
list
of
source,
IP
prefixes
in
a
production
data
plane,
but
for
a
functional
reference
model.
It's
fine
for
dash.
C
F
Yes,
yeah,
so
we
we
in
the
background
and-
and
you
know
we
are
continuing
to
investigate
this
and
doing
quite
a
bit
of
work.
We
also
announced
it
in
ocp
that
we
want
to
have
a
virtual
switch
based
on
p40pdk,
a
Sonic
virtual
switch
Implement,
all
the
site
that,
for
us
the
main
goal
is
to
also
Implement
Dash
and,
as
Andy
was
saying,
the
p.
F
A
architecture
you
know
involves
just
these
things
that
we
are
talking
about
from
the
dash
standpoint
so
interested
in
implementing
and
getting
that
open
source
as
well
so
yeah.
In
the
background
we
are
actually
working
on
this
and-
and
you
know,
we
want
to
discuss
more
at
some
point
with
the
community,
maybe
early
next
year
to
see
this
is
the
features
that
we
have
and
what
else
we
need
to
add
to
start
start
using
it.
G
What
is
what
is
the
current
state
of
bmv2
in
in
the
dash
Community
itself,
in
other
words,
have
been
modified,
anything
from
the
vanilla
bmv2
and
if
we
have
what
are
the
list
of
changes
that
have
really
gone
into
it
right.
So
if
there
is
a
list
of
that,
and
then
you
know,
of
course,
we
know
what
the
backlog
is.
As
you
were
mentioning.
G
A
B
Yeah
I
just
there's
some
collaboration,
but
I
think
maybe
one
possibility
here
is
to
try
to
open
up
the
work.
That's
been
done
and
maybe
there's
other
people
that
can
help
volunteer.
It's
a
little
bit
of
a
cipher
to
me
right
now.
What's
what's
happening
back
there
in
terms
of
the
the
bmv2
itself,
and
you
know
the
compiler.
A
E
G
Right
right,
so
this
this
Pro
probably
begs
the
question
that
we
probably
we
should
keep
some
sort
of
a
document
outlining
within
the
community
on
the
website.
I
guess
on
the
GitHub
that
what
has
been
modified
and
what
is
still
pending
and
so
forth
kind
of
a
thing
right.
So
so
the
people
know
that,
when
they're
testing
with
bmv2
they
there
is
no
surprise
in
terms
of
oh
I.
Have
this
BMV
too?
G
A
A
E
I
think
we
do
need
a
project
owner
and
somebody
who
can
can
explain
what
the
state
of
this
is,
because
it's
even
not
clear
to
me
like,
what's
in,
what's
to
be
done,
you
know
it
seems
like
the
Hardware's
got
ahead
of
the
actual.
You
know
software
simulator.
So
it's
not
clear
that
somebody
is
the
owner
of
that
and
can
articulate
that
clearly
and
then
Marion's
not
going
to
be
the
owner.
Then
we
should
maybe
be
looking
for
an
owner
yeah.
B
Yeah,
my
feeling
is
the
scope
of
this
project
and
the
complexity.
There's
like
a
handful
of
people
who
I
know
in
the
industry
who
can
sort
of
hold
it
all
in
their
head
at
once,
right
and
and
and
are
uniquely
talented,
to
build
it.
They
just
sat
down
and
did
it
full
time,
they'd,
probably
get
done
or
the
equivalent
of
one
person
spread
out,
but
that
that
type
of
knowledge
and
expertise
is
not
widespread
and
it
does
require
like
real
dedication
to
finish
it
and
that's
that's!
B
E
A
E
A
All
the
time
really,
this
is
a
great
question.
Honey,
really
great
question
and
I
agree
with
you.
We
do
have
you
know
a
whole
list
of
items,
but
the
you
know
overall
ownership
and
Status
would
be
great
and
yeah.
It
can
be
a
little
bit
frustrating
when
we
have
tests
that
we're
trying
to
run,
and
we
have
to
mark
them,
as
you
know,
do
not
run
in
bmv2,
so
I
get
it.
A
And
we
were
talking
about
that
a
little
bit
Gerald
before
you
joined
the
calls.
You
know
we
have
test
contributions
to
the
dash
repo
and
you
can
see
here
in
the
pull
requests.
You
know,
they're,
adding
functional
test
cases
adding
different.
You
know
baby
hero
tests,
bgp
test
cases
and
we're
marking
them.
You
know,
does
not
run
in
BMV
too
just
FYI
right,
okay,
great,
so
that's
about
all
I
had
for
today.
Do
you
guys
have
anything
you
want
to
talk
about
before
the
holiday
break.
E
A
E
For
the
second
right,
yeah
I,
don't
think
the
next
two
weeks,
what
we've
been
finding
internally
at
Microsoft
is
day
by
day.
People
are
just
coming
forth
and
say:
I'm
taking
holidays,
so
yeah,
like
everybody,
is
almost
like
literally
almost
every
developer
that
we
have
is
like
not
going
to
be
here
for
the
next
two
weeks,
so
I
suspect
that's
the
same
for
most
companies,
probably.
A
E
A
B
A
I'll
make
sure
Yusuf
is
here
or
I
can
always
pop
in
Helm,
but
Gerald's
right
yeah.
We
have
a
lot
of
people
just
raising
their
hands,
saying
yep
we're
out
too
so.
I
hope
you
guys
have
a
good
break
regardless
and
I'll
send
out
notes.
You
know
this
week,
like
I
always
do
kind
of
outlining
the
two
weeks
will
be
off
and
then
the
one
week
we're
picking
up
on
January,
2nd,
okay,
and
if
you
have
any
ideas,
I
mean
any
creative
ideas
and
I'll
take
it
offline
with
Gerald
too
about
the
whole
bmv2
ownership.